In medical imaging, tissue characterization is fundamental for the assessment of brain tumor pathology and morphology. On T1 or T2-weighted MR images, even with contrast agents, distinction of pathologic tissue from normal tissue often is limited. Excellent contrast, however, results from diffusion weighted MR imaging in the case of acute stroke. Moreover, diffusion tensor imaging permits a detailed characterization of white matter tracts based on the principal diffusion direction and diffusion anisotropy. Imaging with very high diffusion weighting, which can be performed on clinical MR systems, has revealed that the diffusion related MR signal decay in tissues, unlike the diffusion attenuated signal in fluids, is non-monoexponential. The departure from a pure monoexponential signal decay was found to be particularly strong in tumor tissue. A well-founded hypothesis is that the deviation may be linked to restricted diffusion and that it may indeed provide insights into cellular architecture at the micrometer dimension level. Ultimately, at this level of resolution, it is light microscopy analysis of histologic tissue samples that permits definitive differentiation between tumor and normal tissue. The main goal of this research is to further our understanding about restricted diffusion by comparing the MR diffusion measurements with detailed analysis of light-microscopy image data obtained in tissue sections that underwent diffusion scans. In human nerve fiber samples, we would like to investigate the relationship between diffusion anisotropy and nerve fiber density or degree of myelination. In animal brain tumor models, we would like to compare the deviation from a monoexponential MR signal decay with cell density, nucleus to cytoplasma size ratio, or cell size distribution. MR diffusion imaging is widely used for medical diagnosis and is rapidly exploring new indications. This study, therefore, proposes basic research, which will improve the interpretation of clinical MR diffusion image data. [unreadable] [unreadable] [unreadable]